[llvm] [AArch64] Fix SVE scalar fcopysign lowering without neon. (PR #129787)

[David Sherwood via llvm-commits]

================
@@ -10686,6 +10686,35 @@ SDValue AArch64TargetLowering::LowerFCOPYSIGN(SDValue Op,
     return convertFromScalableVector(DAG, VT, Res);
   }
 
+  // With SVE, but without Neon, extend the scalars to scalable vectors and use
+  // a SVE FCOPYSIGN.
+  if (!VT.isVector() && VT.isSimple() && !Subtarget->isNeonAvailable()) {
+    EVT SVT;
+    switch (VT.getSimpleVT().SimpleTy) {
----------------
david-arm wrote:

It might be better to reuse existing code and replace the switch statement with:

```
  EVT SVT = getPackedSVEVectorVT(VT);
```

and then add guards in the if statement for supported types:

```
  if ((VT == MVT::f16 || VT == MVT::f32 || VT == MVT::f64) && !Subtarget->isNeonAvailable())
```

https://github.com/llvm/llvm-project/pull/129787